Damien L-G
Damien L-G
Consider leveraging `_bit_scan_reverse` and similar builtins in `bit_width()`
Collecting here references to recent PRs that propagated the transition from `` * https://github.com/kokkos/kokkos/pull/7952 * https://github.com/kokkos/kokkos/pull/7949 * https://github.com/kokkos/kokkos/pull/7943 * https://github.com/kokkos/kokkos/pull/7934 * https://github.com/kokkos/kokkos/pull/7900 * https://github.com/kokkos/kokkos/pull/7899 * https://github.com/kokkos/kokkos/pull/7848 * https://github.com/kokkos/kokkos/pull/7840 * https://github.com/kokkos/kokkos/pull/7839...
We might want to add coverage for this :/
We do currently define a `KOKKOS_IMPL_ARCH_NVIDIA_GPU` macro that does what you are asking for, but obviously it is considered private. In the desul atomics implementation we define macros to say...
Can you post a link to the discord discussion? If the command to install a C++ library VS CUDA C++ library does not specify the language then I see how...
> #102 should fix those issues. Mixing unrelated things into the same PR makes it difficult to audit changes. You should consider dedicate a PR to getting rid of code...
> > What behavior do we get when/o the changes in the source files? > > The data pointer doesn't change (movnig a pointer doesn't change it's value). Why did...
> > Why did you change it as part of this PR? > > I thought I had pushed these changes before (they are dated from July 11) already when...
> @dalg24 ping I am unconvinced that this is useful. Maybe you could pitch that with the build and packaging working group cc @cedricchevalier19